DISCOVERY OF ULTRA-FAST OUTFLOWS IN A SAMPLE OF BROAD-LINE RADIO GALAXIES OBSERVED WITH SUZAKU

As plausible phenomenological representation of the continuum in 3.5–10.5 keV, we adopt a single power-law model. We did not find it necessary to include neutral absorption from our own Galaxy as the relatively low column densities involved (Dickey & Lockman 1990; Kalberla et al. 2005) have negligible effects in the considered energy band, see Table 1. The only exception is 3C 445, where the continuum is intrinsically absorbed by a column density of neutral/mildly ionized gas as high as N_H ∼ 10²³ cm⁻² (Sambruna et al. 2007); for this source we included also a neutral intrinsic absorption component with a column density of N_H ≃ 2 × 10²³ cm⁻² (see Table 2). A more detailed discussion of absorption in this source using Chandra and Suzaku data is presented in J. N. Reeves et al. (2010, in preparation) and V. Braito et al. (2010, in preparation).

The ratios of the spectral data against the simple (absorbed for 3C 445) power-law continuum for the five BLRGs are shown in the upper panels of Figures 1, 3, and 5. Some additional spectral complexity can be clearly seen, such as an ubiquitous, prominent neutral Fe Kα emission line at the rest-frame energy of 6.4 keV, absorption structures at energies greater than 7 keV (3C 111, 3C 120, and 3C 390.3), and narrow emission features redward (3C 445) and blueward (3C 120 and 3C 382) to the neutral Fe Kα line. To model the emission lines, we added Gaussian components to the power-law model, including the Fe Kα emission line at E ≃ 6.4 keV and ionized Fe K emission lines in the energy range E ∼ 6.4–7 keV, depending on the ionization state of iron, which in this energy interval is expected to range from Fe ii up to Fe xxvi.

Zoom In Zoom Out Reset image size

We find that the baseline model composed by a power-law plus Gaussian Fe K emission lines provides an excellent phenomenological characterization of the 3.5–10.5 keV XIS data with the lowest number of free parameters. The results of the fits for the five BLRGs are reported in Table 2. Note that only those emission lines with detection confidence levels greater than 99% were retained in the following fits. The weak redshifted emission line present in 3C 445 was not included because it has negligible effect on the fit results; this line will be discussed by V. Braito et al. (2010, in preparation).

As apparent from Figures 1, 3, and 5, several absorption dips are present in the residuals of the baseline model in various cases. To quantify their significance, we computed the Δχ² deviations with respect to the baseline model (Section 3.1) over the whole 3.5–10.5 keV interval. The method is similar to the one used by the steppar command in XSPEC to visualize the error contours, but in this case the inner contours indicate higher significance than the outer ones (e.g., Miniutti & Fabian 2006; Miniutti et al. 2007; Cappi et al. 2009; Tombesi et al. 2010).

The analysis has been carried out for each source spectrum as follows: (1) we first fit the 3.5–10.5 keV data with the baseline model and stored the resulting χ² and (2) a further narrow, unresolved (σ = 10 eV) Gaussian test line was then added to the model, with its normalization free to have positive or negative values. Its energy was stepped in the 4–10 keV band at intervals of 100 eV in order to properly sample the XIS energy resolution, each time making a fit and storing the resulting χ² value. In this way we derived a grid of χ² values and then plot the contours with the same Δχ² with respect to the baseline model.

The contour plots for the different sources are reported in the lower panel of Figures 1, 3, and 5. The contours refer to Δχ² levels of −2.3, −4.61, and −9.21, which correspond to F-test confidence levels of 68% (red), 90% (green), and 99% (blue), respectively. The position of the neutral Fe Kα emission line at rest-frame energy E = 6.4 keV is marked by the dotted vertical line. The arrows indicate the position of the blueshifted absorption lines detected at ⩾99%. The black contours indicate the baseline model reference level (Δχ² = +0.5).

We then proceeded to directly fit the spectra, adding Gaussian absorption lines where indications for line-like absorption features with confidence levels greater than 99% were present. As already noted in Section 3.1, we checked that neglecting to include the weak redshifted emission line apparent only in the spectrum of 3C 445 has no effect on the fit results. The detailed fitting and modeling of the Fe K absorption lines is reported in Table 3 and is discussed in Appendix A for each source.

The contour plots in the lower panels of Figures 1, 3, and 5 visualize the presence of spectral structures in the data and simultaneously give an idea of their energy, intensity, and confidence levels using the standard F-test. However, they give only a semi-quantitative indication and the detection of each line must be confirmed by directly fitting the spectra. Moreover, it has been demonstrated that the F-test method can slightly overestimate the actual detection significance for a blind search of emission/absorption lines as it does not take into account the possible range of energies where a line might be expected to occur, nor does it take into account the number of bins (resolution elements) present over that energy range (e.g., Protassov et al. 2002). This problem requires an additional test on the red/blueshifted lines significance and can be solved by determining the unknown underlying statistical distribution by performing extensive Monte Carlo (MC) simulations (e.g., Porquet et al. 2004; Yaqoob & Serlemitsos 2005; Miniutti & Fabian 2006; Markowitz et al. 2006; Cappi et al. 2009; Tombesi et al. 2010).

Therefore, we performed detailed MC simulations to estimate the actual significance of the absorption lines detected at energies greater than 7 keV. We essentially tested the null hypothesis that the spectra were adequately fit by a model that did not include the absorption lines. The simulations have been carried out as follows: (1) we simulated a source spectrum using the fakeit command in XSPEC by assuming the baseline model listed in Table 2 without any absorption lines and with the same exposure as the real data. We subtracted the appropriate background and grouped the data to a minimum of 25 counts per energy bin; (2) we fit the faked spectrum with the baseline model in the 3.5–10.5 keV band, stored the new parameters values, and generated another simulated spectrum as in step 2 but using the refined model. This procedure accounts for the uncertainty in the null hypothesis model itself and is particularly relevant when the original data set is noisy; (3) the newly simulated spectrum was fitted again with the baseline model in the 3.5–10.5 keV and the resultant χ² was stored; (4) then, a further Gaussian line (unresolved, σ = 10 eV) was added to the model, with its normalization initially set to zero and then allowed to freely vary between positive and negative values. To account for the range of energies in which the line could be detected in a blind search, we stepped its centroid energy between 7 keV and 10 keV at intervals of 100 eV to sample the XIS energy resolution, fitting each time and storing only the maximum of the resultant Δχ² values. The procedure was repeated S = 1000 times and consequently a distribution of simulated Δχ² values was generated. The latter indicates the fraction of randomly generated emission/absorption features in the 7–10 keV band that are expected to have a Δχ² greater than a threshold value. In particular, if N of the simulated Δχ² values are greater or equal to the real value, then the estimated detection confidence level from MC simulations is simply 1 − N/S.

The MC detection probabilities for the absorption lines are given in Table 3. The values are in the range of 91% and 99.9%. As expected, these estimates are slightly lower than those derived from the F-test (⩾99%) because they effectively take into account the randomly generated lines in the whole 7–10 keV energy interval.

To model the absorbing material that is photoionized by the nuclear radiation, a grid with the Xstar code (Kallman & Bautista 2001) was generated. We modeled the nuclear X-ray ionizing continuum with a power law with photon index Γ = 2, as usually assumed for Seyfert galaxies, which takes into account the possible steeper soft excess component (e.g., Bianchi et al. 2005). A different choice of the power-law slope in the range Γ = 1.5–2.5 has negligible effects (<5%) on the parameter estimates in the considered Fe K band, E = 3.5–10.5 keV. Moreover, as already noted by McKernan et al. (2003a), the presence or absence of the possible UV-bump in the spectral energy distribution (SED) has a negligible effect on the parameters of the photoionized gas in the Fe K band because in this case the main driver is the ionizing continuum in the hard X-rays (E > 6 keV). Standard solar abundances are assumed throughout (Grevesse et al. 1996).

The velocity broadening of absorption lines from the photoionized absorbers in the central regions of Seyfert galaxies is dominated by the turbulence velocity component, commonly assumed to be in the range ∼100–1000 km s⁻¹ (e.g., Bianchi et al. 2005; Risaliti et al. 2005; Cappi et al. 2009, and references therein). The energy resolution of the XIS instruments in the Fe K band is FWHM ∼ 100–200 eV, implying that lines with velocity broadening lower than ∼2000–4000 km s⁻¹ are unresolved. Therefore, given that we cannot estimate the velocity broadening of the lines directly from the spectral data, we generated an Xstar grid assuming the most likely value for the turbulent velocity of the gas of 500 km s⁻¹. We checked that for higher choices of this parameter, the resultant estimate of the ionization parameter was not affected, although the derived absorber column density was found to be slightly lower. This is due to the fact that the core of the line tends to saturate at higher N_H, upon increasing the velocity broadening (e.g., Bianchi et al. 2005). The opposite happens for lower choices of the turbulent velocity. However, the resulting difference of ∼5%–10% in the derived values is completely negligible and well within the measurement errors.

Therefore, we apply this photoionization grid to directly model the different absorption lines detected in the Fe K band. The free parameters of the model are the absorber column density N_H, the ionization parameter ξ, and the velocity shift v. We let the code find the best-fit values and it turned out that the gas is systematically outflowing, with velocities consistent with those derived from the Gaussian absorption line fits (see Section 3.2 and Appendix A for a detailed discussion of each source). The Xstar parameters are reported in Table 4 and the best-fit models are shown in Figures 2 and 4. A consistency check of the results from a broadband spectral analysis is reported in Section 4.2.

Zoom In Zoom Out Reset image size

In this section, we summarize the results of the spectral fits to the 3.5–10.5 keV XIS-FI spectra of the BLRGs of our sample with a model consisting of the baseline model plus absorption lines and a detailed photoionization grid (see above). Results for individual sources are discussed in Appendix A.

The results of the fits with the baseline model are listed in Table 2, while the residuals of this model are shown in Figures 1, 3, and 5 for the five BLRGs, together with the Δχ² contours. As mentioned above, absorption dips are visible and to assess their statistical significance we used both the F-test and extensive MC simulations. The results of these tests, reported in Table 3, establish that only in 3/5 sources do we reliably detect absorption features at energies ∼ 7.3–7.5 keV and 8.1–8.7 keV, namely, in 3C 111, 3C 120b, and 3C 390.3. In these three sources, the absorption lines are detected with confidence levels higher than 99% with the F-test and higher than 91% with the MC method (Table 3). We fit the absorption features by adding narrow Gaussian components, or a blend of narrow components, to the baseline model. The Gaussian parameters are reported in Table 3.

Zoom In Zoom Out Reset image size

Given the high cosmic abundance of Fe, the most intense spectral features expected from a highly ionized absorber in the 3.5–10.5 keV band are the K-shell resonances of Fe xxv and Fe xxvi (e.g., Kallman et al. 2004). However, the rest-frame energies of the detected absorption lines are in the range ≃7.3–7.5 keV and ≃8.1–8.8 keV, larger than the expected energies of the atomic transitions for Fe xxv and Fe xxvi. An interesting possibility is that the absorption lines detected in the BLRGs are due similarly, to those recently observed in Seyferts, to blueshifted resonant lines of highly ionized Fe, thus implying the presence of fast outflows in radio-loud AGNs as well. If we hold this interpretation true, the derived outflow velocities are in the range ≃0.04–0.15c.

We also performed more physically consistent spectral fits using the Xstar photoionization grid described in Section 3.4 (see Figures 2 and 4). Good fits are obtained with this model, yielding ionization parameters log ξ ≃ 4–5.6 erg s⁻¹ cm and column densities N_H ≃ 10²²–10²³ cm⁻². The derived blueshifted velocities are consistent with those from the simple phenomenological fits, v ≃ 0.04–0.15c (see Table 4). We note that, given the very high ionization level of this absorbing material, no other significant signatures are expected at lower energies as all the elements lighter than iron are almost completely ionized.

Zoom In Zoom Out Reset image size

An important caveat is that the velocities and column densities derived by fitting the spectral data with the Xstar grid depend on the unknown inclination angle of the outflow with respect to the line of sight. In other words, they depend on whether we are actually looking directly down to the outflowing stream or intercepting only part of it (e.g., Elvis 2000). Therefore, the obtained values (see Table 4) are only conservative estimates and represent lower limits.

In conclusion, we detected for the first time in radio-loud AGNs at X-rays, absorption lines in the energy range 7–10 keV in the Suzaku XIS spectra of 3/5 BLRGs—3C 111, 3C 390.3, and 3C 120. If interpreted as blueshifted resonant absorption lines of highly ionized Fe, the features imply the presence of ultra-fast (v ∼ 0.04–0.15c) outflows in the central regions of BLRGs. In Section 5, we discuss more in depth this association and the inferred outflow physical properties.

As a consistency check of the Fe K band (E = 3.5–10.5 keV) based results, we exploited the broadband capabilities of Suzaku combining the XIS and PIN spectra. The energy band covered in this way is very broad, from 0.5 keV up to 50 keV. We downloaded and reduced the PIN data of 3C 111, 3C 390.3, and 3C 120 and analyzed the combined XIS-FI and PIN spectra. For 3C 390.3 and 3C 120, we applied the broadband models already published in the literature by Sambruna et al. (2009) and Kataoka et al. (2007). Instead, for 3C 111, we used the broadband model that will be reported by us in L. Ballo et al. (2010, in preparation). This is essentially composed by a power-law continuum with Galactic absorption, plus cold reflection (R≲1) and the Fe Kα emission line at E ≃ 6.4 keV. The resultant power-law photon index of this fit is Γ ≃ 1.6, which is slightly steeper than the estimate of Γ ≃ 1.5 from the local continuum in the 3.5–10.5 keV band (see Table 2). We included the neutral Galactic absorption component in all broadband fits (see Table 1). Then, we modeled the blueshifted absorption lines with the Xstar photoionization grid already discussed in Section 3.4., letting the column density, ionization parameter, and velocity shift vary as free parameters.

The best-fit estimates of the Fe K absorbers derived from these broadband fits are completely consistent with those reported in Table 4. In particular, for 3C 111 we obtain an ionization parameter of log ξ = 4.9^+0.2_−0.4 erg s⁻¹ cm, a column density of N_H > 1.5 × 10²³ cm⁻², and an outflow velocity of v_out = +0.039 ± 0.003c. For 3C 390.3, we estimate log ξ = 5.6 ± 0.5 erg s⁻¹ cm, N_H > 2 × 10²² cm⁻², and v_out = +0.146 ± 0.007c. Finally, for 3C 120b, we derive log ξ = 3.7 ± 0.2 erg s⁻¹ cm, N_H = (1.5 ±  0.4) × 10²² cm⁻², and v_out = +0.075 ±  0.003c. This also assures that the addition of a weak reflection component with R < 1 (e.g., Sambruna et al. 2009; Kataoka et al. 2007; L. Ballo et al. 2010, in preparation) does not at all change the fit results.

Moreover, it is important to note here that we do not find any evidence for a lower ionization (log ξ ≲ 3 erg s⁻¹ cm) warm absorber at E ≲ 3 keV in these three sources. This rules out any possible systematic contamination from moderately ionized iron and strengthens the interpretation of the absorption lines at E > 7 keV as genuine blueshifted Fe xxv and Fe xxvi K-shell transitions. As already introduced in Section 1, the only object with the detection of a soft X-ray warm absorber in its high energy resolution Chandra HETG (Reeves et al. 2009a) and XMM-Newton RGS (Torresi et al. 2010) spectra is 3C 382. On the other hand, a heavy soft X-ray absorption from neutral/mildly ionized gas with N_H ∼ 10²³ cm⁻² has been reported in the XMM-Newton spectrum of 3C 445 (Sambruna et al. 2007). This result is also confirmed by a Chandra LETG and a Suzaku broadband spectral analysis that will be presented in J. N. Reeves et al. (2010, in preparation) and V. Braito et al. (2010, in preparation), respectively. However, we did not find any significant narrow Fe K absorption line features in the 7–10 keV Suzaku XIS spectra of these two sources from this analysis.

The discovery of UFOs in radio-loud BLRGs parallels the detection of UFOs in the X-ray spectra of several Seyfert galaxies and radio-quiet quasars (e.g., Chartas et al. 2002, 2003; Pounds et al. 2003; Dadina et al. 2005; Markowitz et al. 2006; Braito et al. 2007; Cappi et al. 2009; Reeves et al. 2009b). The presence of UFOs in radio-quiet sources was recently established through a systematic, uniform analysis of the XMM-Newton archive on a large number of sources (Tombesi et al. 2010), overcoming possible publication biases (e.g., Vaughan & Uttley 2008).

While a uniform analysis was also performed in this work, it should be noted that our small sample is not complete and the results might not be representative of the global population of BLRGs. Therefore, to obtain better constraints on the statistical incidence and parameters of UFOs in BLRGs, it is imperative to expand the sample of sources with high-quality X-ray observations in the next few years through Suzaku and XMM-Newton observations of additional sources.

However, it has been claimed that part (or even all) of the blueshifted ionized absorption features detected in the X-ray spectra of bright AGNs could be affected by contamination from local (z ≃ 0) absorption in our Galaxy or by the warm/hot intergalactic medium (WHIM) at intermediate redshifts, due to the fact that some of them have blueshifted velocities comparable to the sources cosmological redshifts (e.g., McKernan et al. 2003b, 2004, 2005). We performed some tests to look into this scenario. We can use the velocity information and compare the absorber blueshifted velocities with the cosmological redshifts of the sources. The blueshifted velocities of the absorbers detected in 3C 120 and 3C 390.3 (see Table 4) are much larger than the sources' cosmological redshifts (see Table 1). This conclusion is strong enough to rule out any contamination due to absorption from local or intermediate redshift material in these two sources. However, the derived blueshifted velocity of v = +0.041 ± 0.003c for the highly ionized absorber in 3C 111 is instead somewhat similar to the source cosmological redshift of z = 0.0485 and needs to be investigated in more detail. The difference between the two values is zc − v ≃ 0.007c, which could indicate absorption from highly ionized material either in our Galaxy and outflowing with that velocity (v ∼ 2000 km s⁻¹) along the line of sight or at rest and located at that intermediate redshift (z ≃ 0.007).

The galaxy 3C 111 is located at a relatively low latitude (b = −88) with respect to the Galactic plane and therefore its X-ray spectrum could be, at some level, affected by local obscuration. However, the estimated column density of Galactic material along the line of sight of the source is N_H ∼ 3 × 10²¹ cm⁻² (Dickey & Lockman 1990; Kalberla et al. 2005), which is far too low to explain the value of N_H ∼ 10²³ cm⁻² of the absorber from fits of the Suzaku spectrum (see Table 4). Nevertheless, the source is located near the direction of the Taurus molecular cloud, which is the nearest large star-forming region in our Galaxy.

A detailed optical and radio study of the cloud has been reported by Ungerer et al. (1985). From the analysis of the emission from the stars in that region and the molecular emission lines, the authors estimated several parameters of the cloud, such as the location at a distance of ∼400 pc (with a linear extent of ∼5 pc), a kinetic temperature of T ≃ 10 K, a typical velocity dispersion of ∼1–3 km s⁻¹, and a low number density of n(H₂) ∼ 300 cm⁻³. These parameters are completely inconsistent with the properties of the X-ray absorber. In fact, the extreme ionization level (log ξ ∼ 5 erg s⁻¹ cm) needed to have sufficient Fe xxvi ions would completely destroy all the molecules and ionize all the lighter atoms. The temperature associated with such photoionized absorber (T ∼ 10⁶ K) is larger than that estimated for the Taurus molecular cloud. Also the outflow velocity of ∼2000 km s⁻¹, expected if associated with such Galactic clouds, would be substantially higher than the velocity dispersion estimated by Ungerer et al. (1985). The authors also stated that the mapping of the visual extinction due to the molecular cloud clearly shows that the region of the cloud in front of 3C 111 is not the densest part (see Figure 3 of Ungerer et al. 1985).

This result is also supported by a recent detailed X-ray study of this region that has been performed by the XMM-Newton Extended Survey of the Taurus Molecular Cloud project (Güdel et al. 2007). This work has been focused on the study of the stars and gas located in the most populated ≃5 deg² region of the Taurus cloud. With a declination of ∼38°, 3C 111 is located outside the edge of this complex region, where mainly only extended cold and low density molecular clouds are distributed (see Figure 1 of Güdel et al. 2007). Therefore, the identification of the highly ionized absorber of 3C 111 with local Galactic absorption is not feasible.

We also find that association with absorption from the WHIM at intermediate redshift (z ∼ 0.007) is very unlikely. In fact, this diffuse gas is expected to be collisionally ionized, instead of being photoionized by the AGN continuum. Therefore, the temperature required to have a substantial He/H-like iron population would be much higher (T ∼ 10⁷–10⁸ K) than the expected T ∼ 10⁵–10⁶ K. The huge column density of gas (N_H ≳ 10²³ cm⁻²) required to reproduce the observed features is also too high compared to those expected for the WHIM (N_H ≲ 10²⁰ cm⁻²). Moreover, the detection of highly ionized absorbers in 3C 120 and 3C 390.3 with blueshifted velocities substantially larger than the relative cosmological redshifts strongly supports the association of the absorber in 3C 111 with a UFO intrinsic to the source.

Similar conclusions were reached by Reeves et al. (2008) concerning the bright quasar PG 1211+143. The X-ray spectrum of this source showed a blueshifted absorption line from highly ionized iron (Pounds et al. 2003; Pounds & Page 2006) with a blueshifted velocity comparable to the cosmological redshift of the source. This led some authors to suggest its possible association with absorption from intervening diffuse material at z ∼ 0 (e.g., McKernan et al. 2004). However, the detection of line variability on a timescale less than 4 yr, suggesting a compact ∼parsec scale absorber, and the extreme parameters of the absorber, e.g., log ξ ∼ 3–4 erg s⁻¹ cm and N_H ∼ 10²²–10²³ cm⁻², led Reeves et al. (2008) to exclude such interpretation. As pointed out by the authors, the evidence of several other radio-quiet AGNs with Fe K absorption with associated blueshifted velocities higher than the relative cosmological redshift suggested that the case of PG 1211+143 was a mere coincidence (see also Tombesi et al. 2010).

We conclude that the evidence for UFOs in BLRGs from the Suzaku data is indeed robust. In the next section, we examine their physical properties in detail.

From the definition of the ionization parameter ξ = L_ion/nr² (Tarter et al. 1969), where n is the average absorber number density and L_ion is the source X-ray ionizing luminosity integrated between 1 Ryd and 1000 Ryd (1 Ryd = 13.6 eV), we can estimate the maximum distance r of the absorber from the central source. The column density of the gas N_H is a function of the density of the material n and the shell thickness Δr: N_H = nΔr. Making the reasonable assumption that the thickness is less than the distance from the source r and combining with the expression for the ionization parameter, we obtain the upper limit r < L_ion/ξN_H. Using the absorption corrected luminosities L_ion ≃ 2.2 × 10⁴⁴ erg s⁻¹, L_ion ≃ 2.3 × 10⁴⁴ erg s⁻¹, and L_ion ≃ 5.1 × 10⁴⁴ erg s⁻¹ directly estimated from the Suzaku data and the ionization parameters and column densities listed in Table 4, we obtain the limits of r < 2 × 10¹⁶ cm (<0.007 pc), r < 10¹⁸ cm (<0.3 pc), and r < 4 × 10¹⁶ cm (<0.01 pc) for 3C 111, 3C 120, and 3C 390.3, respectively. Using the black hole mass estimates of M_BH ∼ 3 × 10⁹ M_☉ for 3C 111 (Marchesini et al. 2004), M_BH ∼ 5 × 10⁷ M_☉ for 3C 120 (Peterson et al. 2004), and M_BH ∼ 3 × 10⁸ M_☉ for 3C 390.3 (Marchesini et al. 2004; Peterson et al. 2004), the previous limits on r correspond to a location for the absorber within a distance of ∼20r_s, ∼7 × 10⁴r_s, and ∼500r_s from the SMBH, respectively. The expected variability timescale of the absorbers from the light crossing time, t ∼ r/c, is t ∼ 600–700 ks (∼7 days) for 3C 111, t ∼ 1 yr for 3C 120, and t ∼ 15–20 days for 3C 390.3, respectively.

A rough estimate of the escape velocity along the radial distance for a Keplerian disk can be derived from the equation v²_esc = 2GM_BH/r, which can be re-written as v_esc = (r_s/r)^1/2c. Therefore, for 3C 111 the escape velocity at the location of ∼20r_s is v_esc ∼ 0.2c, which is larger than the measured outflow velocity of v∼0.041c (see Table 4). This implies that most likely the absorber is actually in the form of a blob of material that would eventually fall back down, possibly onto the accretion disk. For 3C 120, the measured outflow velocity v∼0.076c (see Table 4) is equal to the escape velocity at a distance of ∼200r_s from the black hole. Therefore, if the launching region is further away than this distance, the ejected blob is likely to escape the system. Concerning 3C 390.3, the measured velocity of v∼0.146c (see Table 4) is larger than the escape velocity at ∼500r_s and equals that at a distance of ∼50–60r_s. Therefore, if the blob of material has been ejected from a location between, say, ∼100r_s and ∼500r_s, it has likely enough energy to eventually leave the system.

We can get an idea of the effectiveness of the AGN in producing outflows by comparing their luminosity with the Eddington luminosity, L_Edd ≃ 1.3 × 10³⁸(M_BH/M_☉) erg s⁻¹. Substituting the estimated black hole mass for each source, we have L_Edd ≃ 3.9 × 10⁴⁷ erg s⁻¹ for 3C 111, L_Edd ≃ 6.5 × 10⁴⁵ erg s⁻¹ for 3C 120, and L_Edd ≃ 3.9 × 10⁴⁶ erg s⁻¹ for 3C 390.3, respectively. From the relation L_bol ≃ 10L_ion (e.g., McKernan et al. 2007), the bolometric luminosities of the different sources are L_bol ≃ 2.2 × 10⁴⁵ erg s⁻¹ for 3C 111, L_bol ≃ 2.3 × 10⁴⁵ erg s⁻¹ for 3C 120, and L_bol ≃ 5.1 × 10⁴⁵ erg s⁻¹ for 3C 390.3, respectively. The ratio L_bol/L_Edd is almost negligible for 3C 111 but it is of the order of ∼0.1–0.4 for 3C 120 and 3C 390.3. These two sources are emitting closer to their Eddington limits and therefore are more capable of producing powerful outflows/ejecta that would eventually leave the system (e.g., King & Pounds 2003; King 2010). This supports the conclusions from the estimates on the location of the ejection regions and the comparison of the outflow velocities with respect to the escape velocities.

Moreover, assuming a constant velocity for the outflow and the conservation of the total mass, we can roughly estimate the mass loss rate $\dot{M}_{\rm out}$ associated to the fast outflows, $\dot{M}_{\rm out}=4\pi C r^2 n m_p v$ (e.g., Blustin et al. 2005; McKernan et al. 2007), where v is the outflow velocity, n is the absorber number density, r is the radial distance, m_p is the proton mass, and C ≡ (Ω/4π) is the covering fraction, which in turn depends on the solid angle Ω subtended by the absorber. From the definition of the ionization parameter ξ, we obtain $\dot{M}_{\rm out}=4\pi C \frac{L_{\rm ion}}{\xi } m_p v$. Substituting the relative values, we derive estimates of $\dot{M}_{\rm out} \sim 2\,C$ M_☉ yr⁻¹, $\dot{M}_{\rm out} \sim 17\,C$ M_☉ yr⁻¹, and $\dot{M}_{\rm out} \sim 2\,C$ M_☉ yr⁻¹ for 3C 111, 3C 120, and 3C 390.3, respectively.

The kinetic power carried by the outflows can be estimated as $\dot{E}_K\equiv \frac{1}{2} \dot{M}_{\rm out} v^2$, which roughly corresponds to $\dot{E}_K \sim 4.5\times 10^{43}\,C$ erg s⁻¹, $\dot{E}_K \sim 3\times 10^{45}\,C$ erg s⁻¹, and $\dot{E}_K \sim 1.2\times 10^{45}\,C$ erg s⁻¹ for 3C 111, 3C 120, and 3C 390.3, respectively. Note that, depending on the estimated covering fraction, the kinetic power injected in these outflows can be substantial, possibly reaching significant fractions (∼0.01–0.5) of the bolometric luminosity and can be comparable to the typical jet power of these sources of ∼10⁴⁴–10⁴⁵ erg s⁻¹, the latter being the power deposited in the radio lobes (Rawlings & Saunders 1991).

Therefore, it is important to compare the fraction of the mass that goes into the accretion of the system with respect to that which is lost through these outflows. Following McKernan et al. (2007), we can derive a simple relation for the ratio between the mass outflow rate and the mass accretion rate, i.e., $\dot{M}_{\rm out}/\dot{M}_{\rm acc}\simeq 6000\, C(v_{0.1}/\xi _{100})\eta _{0.1}$, where v_0.1 is the outflow velocity in units of 0.1c, ξ₁₀₀ is the ionization parameter in units of 100 erg s⁻¹ cm, and η = η_0.1 × 0.1 is the accretion efficiency. Substituting the parameters with their relative values listed in Table 4, we obtain $\dot{M}_{\rm out}/\dot{M}_{\rm acc} \sim 2\,C$ for 3C 111, $\dot{M}_{\rm out}/\dot{M}_{\rm acc} \sim 40\,C$ for 3C 120, and $\dot{M}_{\rm out}/\dot{M}_{\rm acc} \sim 2\,C$ for 3C 390.3, respectively.

These estimates depend on the unknown value of the covering fraction C. A very rough estimate of the global covering fraction of these absorbers can be derived from the fraction of sources of our small sample: C ≃ f = 3/5 ∼ 0.6 (e.g., Crenshaw et al. 1999). This suggests that the geometrical distribution of the absorbing material is not very collimated but large opening angles are favored. The rough estimate C ∼ 0.6 implies the possibility of reaching ratios of about unity or higher between the mass outflow and accretion rates. This means that these outflows can potentially generate significant mass and energy losses from the system. However, the covering fraction crude estimate of C ∼ 0.6 has been derived from a very small sample which is far from being complete and therefore could not be fully representative of the whole population of BLRGs.

The physical characteristics of UFOs here derived for the three BLRGs strongly point toward an association with winds/outflows from the inner regions of the putative accretion disk. In fact, simulations of accretion disks in AGNs ubiquitously predict the generation of mass outflows. For instance, the location, geometry, column densities, ionization, and velocities of our detected UFOs are in good agreement with the AGN accretion disk wind model of Proga & Kallman (2004). In this particular model, the wind is driven by radiation pressure from the accretion disk and the opacity is essentially provided by UV lines. Depending on the angle with respect to the polar axis, three main wind components can be identified: a hot, low density and extremely ionized flow in the polar region; a dense, warm and fast equatorial outflow from the disk; and a transition zone in which the disk outflow is hot and struggles to escape the system. The ionization state of the wind decreases from polar to equatorial regions. Instead, the column densities increase from polar to equatorial, up to very Compton-thick values (N_H > 10²⁴ cm⁻²). The outflows can easily reach large velocities, even higher than ∼10⁴ km s⁻¹.

Lines of sight through the transition region of the simulated outflow, where the density is moderately high (n ∼ 10⁸–10¹⁰ cm⁻³) and the column density can reach values up to N_H ∼ 10²⁴ cm⁻², result in spectra that have considerable absorption features from ionized species imprinted in the X-ray spectrum, mostly with intermediate/high ionization parameters, log ξ ∼ 3–5 erg s⁻¹ cm. This strongly suggests that the absorption material could be observed in the spectrum through Fe K-shell absorption lines from Fe xxv and Fe xxvi (e.g., Sim et al. 2008; Schurch et al. 2009; Sim et al. 2010), in complete agreement with our detection of UFOs. In particular, Sim et al. (2008) and Sim et al. (2010) used their accretion disk wind model to successfully reproduce the 2–10 keV spectra of two bright radio-quiet AGNs in which strong blueshifted Fe K absorption lines were detected in their XMM-Newton spectra, namely, Mrk 766 (from Miller et al. 2007) and PG 1211+143 (from Pounds et al. 2003). Notably, the authors have been able to account for both emission and absorption features in a physically self-consistent way and demonstrated that accretion disk winds/outflows might well imprint also other spectral signatures in the X-ray spectra of AGNs (e.g., Pounds & Reeves 2009 and references therein).

Hydrodynamic wind simulations are highly inhomogeneous in density, column and ionization and have strong rotational velocity components. Therefore the outflow, especially in its innermost regions, is rather unstable. In particular, the outflow properties through the transition region show considerable variability and this is expected to be reflected by the spectral features associated with this region, i.e., by the corresponding blueshifted Fe xxv/xxvi K-shell absorption lines.

Proga & Kallman (2004) and Schurch et al. (2009) state that it is possible that some parts or blobs of the flow, especially in the innermost regions, do not have enough power to allow a "true" wind to be generated. In these cases, a considerable amount of material is driven to large-scale heights above the disk but the velocity of the material is insufficient for it to escape the system and it will eventually fall back onto the disk. Despite returning to the accretion disk at larger radii, while it is above the disk, this material can imprint features on the observed X-ray spectrum (e.g., Dadina et al. 2005 and references therein). This can indeed be the case for some of the UFOs discussed here (i.e., 3C 111).

This overall picture is also partially in agreement with what was predicted by the "aborted jet" model by Ghisellini et al. (2004). This model was actually proposed to explain, at least in part, the high-energy emission in radio-quiet quasars and Seyfert galaxies. It postulates that outflows and jets are produced by every black hole accretion system. Blobs of material can then be ejected intermittently and can sometimes only travel for a short radial distance and eventually fall back, colliding with others that are approaching. Therefore, the flow can manifest itself as erratic high-velocity ejections of gas from the inner disk and it is expected that some outflows/blobs are not fast enough to escape the system and will eventually fall back onto the disk. An intriguing possibility could be that these outflows are generated by localized ejection of material from the outer regions of a bubbling corona, which emits the bulk of the X-ray radiation (Haardt & Maraschi 1991), analogous to what has been observed in the solar corona during the coronal mass ejection events (e.g., Low 1996). The velocity and frequency of these strong events should then be limited to some extent, in order not to cause the disruption or evaporation of the corona itself. Such extreme phenomena could then be the signatures of the turbulent environment close to the SMBH.

The detection of UFOs in both radio-quiet and radio-loud galaxies suggests a similarity of their central engines and demonstrates that the presence of strong relativistic jets do not exclude the existence of winds/outflows from the putative accretion disk. Moreover, it has been demonstrated by Torresi et al. (2010) and Reeves et al. (2009a) that a warm absorber is indeed present in BLRGs also (in particular 3C 382) and this indicates that jets and slower winds/outflows can coexist in the same source, even beyond the broad-line region.

However, BLRGs are radio-loud galaxies and they have powerful jets. Therefore, the fact that for BLRGs we are observing down to the outflowing stream at intermediate angles to the jet (∼15°–30°; e.g., Eracleous & Halpern 1998) suggests that the fast winds/outflows we observe are at greater inclination angles with respect to the jet axis, somewhat similar to what is expected for accretion disk winds (e.g., Proga & Kallman 2004). These outflows would then not be able to undergo the processes that instead accelerate the jet particles to velocities close to the speed of light.

For instance, studies of Galactic stellar-mass black holes, or micro-quasars, showed that wind formation occurs in competition with jets, i.e., winds carry away matter halting their flow into jets (e.g., Neilsen & Lee 2009). Given the well-known analogy between micro-quasars and their super-massive relatives, one would naively expect a similar relationship for radio-loud AGNs. The BLRGs 3C 111 and 3C 120 are regularly monitored in the radio and X-ray bands with the VLBA and RXTE as part of a project aimed at studying the disk–jet connection (e.g., Marscher et al. 2002). We have detected UFOs in both of these sources (see Table 4), and indeed in both cases the 4–10 keV fluxes measured with Suzaku corresponded to historical low(est) states if compared to the RXTE long-term light curves. For instance, correlated spectroscopic observations of 3C 111, where the shortest variability timescales are predicted (t ∼ 7 days), during low and high jet continuum states could provide, in a manner analogous to micro-quasars, valuable information on the synergy among disk, jet, and outflows, and go a long way toward elucidating the physics of accretion/ejection in radio-loud AGNs.

However, whether it is possible to accelerate such UFOs with velocities up to ∼0.15c only through UV line-driving is unclear. Moreover, the material needs to be shielded from the high X-ray ionizing flux in the inner regions of AGNs, otherwise it would become overionized and the efficiency of this process would be drastically reduced. Other mechanisms as well are capable of accelerating winds from accretion disks, in particular radiation pressure through Thomson scattering and magnetic forces.

In fact, Ohsuga et al. (2009) proposed a unified model of inflow/outflow from accretion disks in AGNs based on radiation-MHD simulations. Disk outflows with helical magnetic fields, which are driven either by radiation-pressure force or magnetic pressure, are ubiquitous in their simulations. In particular, in their case A (see their Figure 1) a geometrically thick, very luminous disk forms with a luminosity L ∼ L_Edd, which effectively drives a fast Compton-thick wind with velocities up to ∼0.2–0.3c. It is important to note that the models of Ohsuga et al. (2009) include both radiation and magnetic forces which, depending on the state of the system, can generate both relativistic jets and disk winds.

Moreover, King & Pounds (2003) and King (2010) showed that black holes accreting with modest Eddington rates are likely to produce fast Compton-thick winds. They considered only radiation pressure and therefore fast winds can be effectively generated by low magnetized accretion disks as well. In particular, King (2010) derived that Eddington winds from AGNs are likely to have velocities of ∼0.1c and to be highly ionized, showing the presence of helium- or hydrogen-like iron. These properties strongly point toward an association of our detected UFOs from the innermost regions of AGNs with Eddington winds/outflows from the putative accretion disk.

Depending on the estimated covering fraction, the derived mass outflow rate of the UFOs can be comparable to the accretion rate and their kinetic power can correspond to a significant fraction of the bolometric luminosity and is comparable to the jet power. Therefore, the UFOs may have the possibility of bringing significant amount of mass and energy outward, potentially contributing to the expected feedback from the AGN. In particular, King (2010) demonstrated that fast outflows driven by black holes in AGNs can explain important connections between the SMBH and the host galaxy, such as the observed M_BH–σ relation (e.g., Ferrarese & Merritt 2000). These UFOs can potentially provide an even more important contribution to the expected feedback between the AGN and the host galaxy than the jets in radio-loud sources. In fact, even if jets are highly energetic, they are also extremely collimated and carry a negligible mass. Fast winds/outflows from the accretion disks, instead, are found to be massive and extend over wide angles. Thus, we suggest that UFOs in radio-loud AGNs are a new, important ingredient for feedback models involving these sources.

The 3.5–10.5 keV XIS-FI spectrum of 3C 111 is described well by a simple power-law continuum (with Γ ≃ 1.5) and a narrow neutral Fe Kα emission line at the rest-frame energy of 6.4 keV (see Table 2). A detailed broadband spectral analysis of the Suzaku spectrum of this source will be reported in L. Ballo et al. (2010, in preparation). However, as it can be seen from the ratio of the spectrum against a simple power-law continuum reported in the upper panel of Figure 1 (left), further complexities are present in the spectrum. In fact, besides the narrow emission line, two absorption features can be clearly seen at the observed energies of ∼7 keV and ∼8–9 keV. These absorption features are still present in the energy-intensity contour plot (see lower panel of Figure 1, left), which suggest that their detection confidence levels should be higher than 99%.

Therefore, we directly fit the data, adding two further absorption lines to the baseline model. The detailed line parameters are reported in Table 3. The first absorption line is not resolved and is detected at a rest-frame energy of E = 7.26 ± 0.03 keV, with an equivalent width of EW = −31 ± 15 eV. Its detection confidence level is high: 99.9% from the standard F-test and 99% from extensive MC simulations (see Section 3.3). The most intense spectral features expected at energies ≳7 keV are the inner K-shell resonances from Fe xxvi (e.g., Kallman et al. 2004). These lines are those of the Lyman series, that is: the Lyα (1s–2p) at E = 6.966 keV, the Lyβ (1s–3p) at E = 8.250 keV, the Lyγ (1s–4p) at E = 8.700 keV, and the Lyδ (1s–5p) at E = 8.909 keV (all line parameters have been taken from the NIST⁸ atomic database, unless otherwise stated). However, the observed line energy is not consistent with any of these known atomic transitions. If identified with Fe xxvi Lyα resonant absorption, the centroid of the line indicates a substantial blueshifted velocity of +0.041 ± 0.003c.

The second absorption line is at a measured rest-frame energy of E = 8.69^+0.13_−0.08 keV. It is broader than the first one, with a resolved width of σ = 390⁺²⁷⁰₋₇₀ eV and an equivalent width of EW = −154 ± 80 eV (see Table 3). The detection confidence level of the line is higher than 99.9% with both the F-test and MC simulations (see Section 3.3). Also in this case the energy of the line is not consistent with any known atomic transition. If identified with Fe xxvi Lyβ, the centroid of the line indicates a blueshifted velocity of ∼0.05c. This value is comparable with that of the former line. However, if this is the case, the ratio of the EWs of the Fe xxvi Lyα and Lyβ would be ∼0.2. This is at odds with what is expected from theory. In fact, the ratio between these lines must be instead equal to ≃5 (which is the ratio of their oscillator strengths: 0.42 and 0.08, respectively) and it could decrease to a minimum of ≃1 when the lines are substantially saturated (e.g., Risaliti et al. 2005). This would suggest that the second broad absorption line could actually be a blend of different blueshifted resonance lines, such as the Lyβ, Lyγ, and Lyδ. This scenario is supported by the fact that the energy resolution of the XIS instruments degrades with increasing energy (at E ∼ 8–9 keV it is of the order of FWHM ≳ 200 eV) and therefore these lines could not be separated properly.

To test whether a line blend is consistent with the data, we performed a fit adding to the baseline model four additional narrow absorption lines with energies fixed to the expected values for the Fe xxvi Lyman series and leaving their common energy shift as a free parameter. These lines provide a very good modeling of both absorption features at E > 7 keV, with a global Δχ² = 42 for five additional parameters. The probability of having these four absorption lines at these exact energies simply from random fluctuations is very low, about 10⁻⁸. Interestingly enough, their common blueshifted velocity is +0.041 ± 0.004c, consistent with the one calculated above for the first absorption line. The resultant EWs of the four Fe xxvi lines are EW = −25 ± 8 eV for the Lyα, EW = −35 ± 14 eV for the Lyβ, EW = −27 ± −16 for the Lyγ, and EW > − 60 eV for the Lyδ. Their ratios are now consistent with the theoretical expectations and the fact that they are close to unity suggests possible saturation effects.

In order to have a more physically consistent modeling of these spectral features, we performed a fit using the Xstar photoionization grid discussed in Section 3.4. The best-fit parameters are reported in Table 4. We obtained a good fit with a highly ionized absorber (Δχ² = 22 for three additional parameters, required at a level of >99.9%) with an ionization parameter of log ξ = 5.0 ± 0.3 erg s⁻¹ cm and a column density of N_H > 2 × 10²³ cm⁻². The blueshifted velocity is +0.041 ± 0.003c, completely consistent with the value determined above fitting with four simple inverted Gaussian lines. Given the extremely high ionization level of this absorbing material, no other signatures are expected at lower energies as all the elements lighter than iron are fully ionized, as indeed observed.

We conclude that the detected absorption features are actually due to blueshifted Fe xxvi Lyman series lines. In Figure 2 (left), we plot the baseline model composed by a power-law continuum and a neutral Fe Kα emission line, and superimposed on it the model of the absorption features with two simple inverted Gaussian lines and the Xstar model. The plot shows that the two models are almost completely coincident up to the first absorption line (identified as Fe xxvi Lyα at the observed energy of ∼7 keV) and clearly demonstrates that the apparent broadening of the second absorption feature is actually due to a blend of the three higher order Lyman series lines (i.e., Lyβ, Lyγ, and Lyδ).

A consistency check of the Fe K absorber parameters from a broadband XIS+PIN fit is reported in Section 4.2. Instead, the XIS-FI background analysis and the consistency check of the line parameters among the different XIS cameras (see Table 5), along with the discussion of possible alternative modelings of the lines, are reported in Appendix B.

A power-law continuum (with Γ ≃ 1.6) plus a narrow neutral Fe Kα emission line at 6.4 keV provide a good modeling of the 3.5–10.5 keV XIS-FI spectrum of 3C 390.3 (see Table 2). The broadband spectral analysis of this Suzaku data set has been reported by Sambruna et al. (2009). From the spectral ratios and the energy–intensity contour plots of Figure 1 (right), there is indication of a possible narrow absorption feature at the observed energy of ∼7.7 keV, with a detection confidence level greater than 99%.

Therefore, we fit the spectrum adding a further narrow (unresolved) absorption line to the baseline model. The rest-frame energy of the line is E = 8.11 ± 0.07 keV and its equivalent width is EW = −32 ± 16 eV (see Table 3). The detection confidence level of the line is high: 99.9% from the standard F-test and 99.5% from extensive MC simulations (see Section 3.3). Also in this case the energy of the line is not consistent with any known atomic transition. However, the most intense lines expected from a highly ionized absorber at E ≳ 7 keV are the Fe xxvi Lyman series (see Appendix A.1). If identified with Fe xxvi Lyα resonant absorption, the centroid of the line indicates a substantial blueshifted velocity of +0.150 ± 0.005c. In order to derive a more physically consistent modeling of this absorption line, we performed a fit using the Xstar photoionization grid discussed in Section 3.4. The best-fit parameters are reported in Table 4. We obtained a good fit with a highly ionized absorber (Δχ² = 14 for three more parameters, required at the ≃99.5% level) with an ionization parameter of log ξ = 5.6^+0.2_−0.8 erg s⁻¹ cm and a column density of N_H > 3 × 10²² cm⁻². The blueshifted velocity of the absorber is +0.146 ± 0.004c, completely consistent with what was derived from fitting with a simple inverted Gaussian line. Given the very high ionization level of this absorbing material, no other significant signatures are expected at lower energies as all the elements lighter than iron are completely ionized.

The comparison of the best-fit results for 3C 390.3 including the baseline model and upon superimposing the modeling of the blueshifted absorption line (identified as Fe xxvi Lyα) with a simple narrow inverted Gaussian or with the Xstar photoionization code is shown in Figure 2 (right). The two models coincide completely, apart from a few weak higher order Lyman series resonances, which cannot be detected with sufficient significance given the quality of the spectral data.

In Section 4.2, we report a consistency check of the results performing also a broadband XIS+PIN spectral analysis. In Appendix B, we discuss the XIS-FI background analysis, the consistency check of the blueshifted Fe K absorption lines among the different XIS cameras (see Table 5) and also possible alternative modelings.

The 3.5–10.5 keV XIS-FI spectrum of observation 3C 120a (see Section 2) is well modeled by a power-law continuum (Γ ≃ 1.75) and a narrow neutral Fe Kα emission line at the rest-frame energy of 6.4 keV (see Table 2). As it can be seen from the ratio of the spectrum against a power-law continuum and the contour plots in the left part of Figure 3 (upper and lower panels, respectively), there are no significant emission/absorption features in the Fe K band, apart from the narrow Fe Kα emission line.

The 3.5–10.5 keV XIS-FI spectrum of observation 3C 120b (see Section 2) is described well by a power-law continuum (with Γ ≃ 1.6) plus a narrow neutral Fe Kα emission line at E ≃ 6.4 keV and a further narrow emission line at E ≃ 6.9 keV (see Table 2). These overall results are in agreement with the spectral analysis of this data set previously reported by Kataoka et al. (2007). However, the spectral ratio and the energy-intensity contour plots in the right part of Figure 3 (upper and lower panels, respectively), suggest that further complexities might be present in the Fe K band. In particular, there is evidence for absorption structures at the observed energies of ∼7–7.4 keV and ∼8–9 keV. The contours in the right part of Figure 3 (lower panel) suggest that their detection confidence levels are higher than 99%.

A direct spectral fitting revealed that the absorption structures at ∼7–7.4 keV are actually composed of two narrow (unresolved, σ = 10 eV) absorption lines. They are detected at rest-frame energies of E = 7.25 ± 0.03 keV and E = 7.54 ± 0.04 keV, respectively. Their equivalent widths are EW = −10 ± 5 eV and EW = −12 ± 6 eV, respectively. Their detection confidence level is ≃99% from the F-test, which slightly reduces to 91% and 92% from MC simulations (see Section 3.3). The detailed line parameters are listed in Table 3. Their energies are not consistent with any known atomic transition. However, their location in the spectrum and their energy spacing suggest a possible identification with blueshifted resonance absorption lines from Fe xxv Heα (1s²–1s2p) at E = 6.697 keV and Fe xxvi Lyα (1s–2p) at E = 6.966 keV. Their corresponding blueshifted velocities are substantial and consistent with each other, i.e., +0.076 ± 0.003c and +0.076 ± 0.004c, respectively.

The second absorption structure that is observed at the energy of E ∼ 8–9 keV is broad. If modeled with a simple inverted Gaussian, the resultant rest-frame energy is E = 8.76 ± 0.12 keV, with a broadening of σ = 360⁺¹⁶⁰₋₁₂₀ eV and equivalent width of EW = −50 ± 13 eV. Its detection confidence level is 99.9% from the F-test and slightly reduces to 99.8% with MC simulations (see Table 3). Also in this case the energy of the line is not consistent with any known atomic transition. However, from the identification of two previous absorption lines, we can infer the possible presence of other resonance features from the same ionic species. In fact, the lower energy resolution of the instrument at those energies (FWHM ≳ 200 eV) and the spacing with respect to the first two lines suggest this broad absorption structure could actually be a blend of at least two further narrow resonance lines, namely, Fe xxv Heβ (1s²–1s3p) at E = 7.88 keV and Fe xxvi Lyβ (1s–3p) at E = 8.25 keV.

To test the consistency of this global line identification, we performed a fit adding to the baseline model four narrow absorption lines with energies fixed to the expected values for these Fe xxv and Fe xxvi resonances and leaving their common energy shift as a free parameter. This provided a very good modeling of all the absorption structures at E ≳ 7 keV, with a χ² improvement of 25 (for five additional parameters). The global probability to have these four absorption lines at these exact energies simply from random fluctuations is low, about 4 × 10⁻⁴. Interestingly, their common blueshifted velocity is +0.076 ±  0.003, completely consistent with what derived fitting each line separately. The resultant EWs of these lines are EW = −10 ± 5 eV for the Fe xxv Heα, EW = −11 ± 8 eV for the Fe xxv Heβ, EW = −11 ± 7 eV for the Fe xxvi Lyα, and EW = −13 ± 9 eV for the Fe xxvi Lyβ. Their relative ratios are of the order of unity, which would suggest possible saturation effects.

Finally, we performed a fit using the Xstar photoionization grid discussed in Section 3.2 in order to have a more physically consistent modeling of these spectral features. The best-fit parameters are reported in Table 4. We obtained a good fit with a highly ionized absorber (Δχ² = 12 for three more parameters, required at a level of ≃99%) with an ionization parameter of log ξ = 3.8 ± 0.2 erg s⁻¹ cm and a total column density of N_H = 1.1^+0.5_−0.4 × 10²² cm⁻². This model simultaneously takes into account all the four absorption features we discussed previously. The resultant blueshifted velocity is +0.076 ±  0.003c, completely consistent with the value estimated by fitting the lines with simple inverted Gaussians. Given the high ionization level of this absorbing material, no other significant signatures are expected at lower energies as all the elements lighter than iron are almost completely ionized.

The conclusion that the detected absorption features are actually due to Fe xxv and Fe xxvi resonant lines is represented well in Figure 4. Here, we can see the best-fit baseline model composed of a power-law continuum and a neutral Fe Kα emission line and the superimposed modeling of the absorption structures with two narrow and one broad inverted Gaussians or with the physically self-consistent Xstar photoionization code. The plot shows that the two models are completely coincident up to the first two narrow absorption lines (identified as blueshifted Fe xxv Heα and Fe xxvi Lyα) and clearly demonstrates that the broad absorption structure at the higher energy is actually composed by several narrow resonant lines from the same ionic species (i.e., mainly Fe xxv Heβ and Fe xxvi Lyβ) which appear to be blended together due to the lower instrumental resolution and signal to noise in this energy band (this is similar to the conclusion drawn for 3C 111 in Appendix A.1).

We also checked for variability of the blueshifted Fe K absorption lines between observations 3C 120a and 3C 120b. We added three absorption lines to the baseline model of observation 3C 120a, with energies and widths fixed to those of observation 3C 120b, and calculated the 90% lower limits on the EWs. The values are reported in Table 3. Unfortunately, the lower S/N in observation 3C 120a alone does not allow us to affirm that the lack of absorption lines in this observation was due to temporal variability.

The consistency of the results from a broadband XIS+PIN spectral analysis is presented in Section 4.2. Instead, in Appendix B we discuss the XIS-FI instrumental background, the consistency of the blueshifted absorption lines parameters among the different XIS cameras (see Table 5) and their possible alternative identifications.

The 3.5–10.5 keV XIS-FI spectrum of 3C 382 is well represented by a power-law continuum (with Γ ≃ 1.75) plus a narrow neutral Fe Kα emission line at E = 6.4 keV and a further weak narrow emission line at E ≃ 6.9 keV (see Table 2). A broadband spectral analysis of this Suzaku data set will be reported in L. Ballo et al. (2010, in preparation). From the spectral ratio and the energy-intensity contour plots of Figure 5 (left panel), it can be seen that an additional narrow weak emission line at the rest-frame energy of ∼7 keV is observable (we refer the reader to L. Ballo et al. 2010, in preparation). However, there are no significant absorption structures at energies ≳7 keV. We estimated the lower limit for the presence of a narrow blueshifted absorption line at the indicative energy of 8 keV to be EW > − 20 eV (see Table 3).

Zoom In Zoom Out Reset image size

The 3.5–10.5 keV XIS-FI spectrum of 3C 445 is affected by substantial absorption by neutral/mildly ionized material intrinsic to the AGN. The baseline model is composed by a power-law continuum (with Γ ≃ 1.6) absorbed by neutral material (N_H ≃ 2 × 10²³ cm⁻²) and a narrow neutral Fe Kα emission line at 6.4 keV (see Table 2). The broadband spectral analysis of this Suzaku data set will be reported in V. Braito et al. (2010, in preparation). From the spectral ratios and the energy–intensity contour plots of Figure 5 (right panel), there is indication for a possible narrow weak emission feature redward to the Fe Kα line (we refer the reader to V. Braito et al. 2010, in preparation). We estimated the lower limit for the presence of a narrow blueshifted absorption line at the indicative energy of 8 keV to be EW> − 45 eV (see Table 3).